A typical oscillator comprises a resonator (usually, but not always an inductor and a capacitor), an amplifier, a limiter and a load. The resonator is driven by the amplifier, which supplies the resonator with energy. Part of the energy in the resonator is applied to the input of the amplifier through a feedback circuit in order to keep the resonator in oscillation. The resonator and the amplifier thus form a closed loop The rest of the energy is applied to the load, for useful work.
The output of the amplifier is greater than the sum of the power dissipated in the resonator, the power applied to the amplifier, and the power applied to the load. This is because all oscillators require a loop gain greater than unity in order to operate. An oscillator will not function if, at the oscillator frequency, the magnitude of the product of the gain of the amplifier and the feedback factor of the feedback circuit is less than unity. If the loop gain is less than unity, the oscillations will decay to zero. In theory, the oscillator will operate if the loop gain is exactly equal to one, the so-called Barkhausen condition. However, as a practical matter, an oscillator in which the loop gain is exactly unity is an abstraction which is completely unrealizable in practice. Therefore, a practical oscillator always has a loop gain slightly larger than one to ensure that, with incidental yet inevitable variations in circuit parameters, the loop gain does not fall below unity.
Of course, with a loop gain greater than unity, a signal of one volt, for example, appearing initially at the input will, after a trip around the loop, appear at the input as a signal larger than one volt. After another trip around the loop, this larger signal will become still larger, and so on. The amplitude of the oscillator output will therefore continue to increase unless it is limited, either by a discrete limiting circuit or by non-linearities (either inherent or intentionally introduced) in the amplifier or the resonator themselves.
An example of an oscillator in which non-linearities in the resonator are relied on to limit the oscillations is disclosed in U.S. Pat. No. 4,901,038. However, that patent recognizes that the non-linear resonator introduces noise into the oscillator. In particular, low frequency noise in the resonator is converted to high-frequency noise, in particular phase noise. To minimize noise introduced by the non-linear resonator, it is composed of superconducting elements kept at cryogenic temperatures.
There are other noise sources, however, which U.S. Pat. No. 4,901,038 does not take into account and, therefore, does not suggest how to eliminate. There is residual Johnson noise (i.e., thermal noise) even at cryogenic temperatures, and possibly granular noise, in the resonator, which results in noise energy across the spectrum from DC to well above the resonator frequency. In addition, the input circuit to the linear amplifier feeds back both 1/f noise and Johnson noise to the resonator. The 1/f noise is inherent in active semiconductor devices. Below a certain frequency, perhaps 1 MHz, noise usually increases with decreasing frequency, approximately proportional to 1/f. This 1/f noise is usually attributed to surface conduction and modulation effects in the semiconductor device. In the medium frequency range, the noise figure is constant and lowest for a given device. At higher frequencies, the noise begins to increase again with frequency. Moreover, even a good "linear" amplifier will have some residual non-linearity which frequency-mixes, or beats with, any noise energy coming into the amplifier. The load also feeds back Johnson noise to the resonator. To provide power to an output load requires more energy (i.e., higher gain) from the linear amplifier, which requires additional amplifier current which, in turn, produces more noise. Finally, a separate limiter introduces noise due to its limiting action. This can include both 1/f noise and broadband Johnson noise.
All of these noise sources are known to contribute to both amplitude noise (in phase with the oscillator signal) and phase noise (out of phase with the oscillator signal). Phase noise is especially detrimental, in that it introduces spurious frequency components which limit the usefulness of the oscillator in many applications.
It is therefore one object of the invention to provide an oscillator which provides a minimum of noise fed back into the resonator, which, in turn, provides an oscillator with extremely low noise and, in particular, low phase noise in the vicinity of the oscillator center frequency.